The application of computer-assisted methodology to investigate structure-property relationships for polyhalogenated biphenyls are described. Multiple linear regression methods were used to correlate the properties of these compounds with descriptors encoding structural characteristics of the molecules. In the first study, gas chromatographic retention times of polychlorinated bipenyls (PCBs) were predicted by using simple molecular structure descriptors. Two excellent equations with R2 greater than 0.99 were developed. Descriptors representing degree of chlorination, substitution pattern, and the geometry of the molecules were included in the equations. The minimum number of compounds needed to predict at retention times of all 209 PCBs was also investiged. About 40 compounds are needed to accurately predict the rentntion of all PCB congeners. the importance of obtaining representative samplings of the PCBs was also noted. The methodology was also applied to the prediction of Henry's law constants (HLC) pf PCBs. HLCs were predicted using two methods. The first method involved development of a single equation which directly related HLCs with molecular structures, while the second method involved a two-step procedure in which the solubility and the vapor pressure were first predicted and HLCs were obtained from the ratio of two quantities. Using the firstmethod, a five-varible equation containing deswcriptors which encode the size, shape and polarizability of the molecules were develoepd. It was observed thea HLCs calculated from the single equation were closer to experimental values than those calculated from the ratio of vapor pressure and solubility. The third study involved prediction of gas and liquid chromatographic retention indices of polyhalogenated biphenyls. Excellent equations with high degree of statistical significance were develoepd for both data sets. For gas chromatography, descriptors encoding the size and general shape of the molecules were found important in predicting thier rettention indices, while for liquid chromatography, charge-weighted surface areas were among the important descriptors in the models. Probably because more interactions were involved, the models for lC were not as good as those for GC.